Adjuvant transarterial chemoembolization after curative resection of hepatocellular carcinoma: Propensity score analysis

Abstract

AIM: To compare survival and recurrence in hepatocellular carcinoma (HCC) patients who did or did not receive adjuvant transarterial chemoembolization (TACE).

METHODS: A consecutive sample of 229 patients who underwent curative resection between March 2007 and March 2010 in our hospital was included. Of these 229 patients, 91 (39.7%) underwent curative resection followed by adjuvant TACE and 138 (60.3%) underwent curative resection alone. In order to minimize confounds due to baseline differences between the two patient groups, comparisons were conducted between propensity score-matched patients. Survival data and recurrence rates were compared using the Kaplan-Meier method. Independent predictors of overall survival and recurrence were identified using Cox proportional hazard regression.

RESULTS: Among 61 pairs of propensity score-matched patients, the 1-, 2-, and 3-year overall survival rates were 95.1%, 86.7%, and 76.4% in the TACE group and 86.9%, 78.5%, and 73.2% in the control group, respectively. At the same time, the TACE and control groups also showed similar recurrence rates at 1 year (13.4% vs 24.8%), 2 years (30.6% vs 32.1%), and 3 years (40.1% vs 34.0%). Multivariate Cox regression identified serum alpha-fetoprotein level ≥ 400 ng/mL and tumor size > 5 cm as independent risk factors of mortality (P < 0.05).

CONCLUSION: As postoperative adjuvant TACE does not improve overall survival or reduce recurrence in HCC patients, further study is needed to clarify its clinical benefit.

Core tip: This study examined survival and recurrence in hepatocellular carcinoma patients who did or did not undergo transarterial chemoembolization (TACE). Analyses were performed with all patients and with propensity score-matched pairs. Both analyses suggest adjuvant TACE does not reduce recurrence. Furthermore, adjuvant TACE does not improve overall survival among propensity score-matched patients. These findings raise important questions about the efficacy of postoperative adjuvant TACE that future work should address in order to justify the use of this technique as well as identify the patient populations most likely to benefit from it.

INTRODUCTION

Hepatocellular carcinoma (HCC) is the fifth most common cancer in men and the seventh most common in women, and it ranks as the third leading cause of cancer-related death worldwide[1,2]. HCC is difficult to diagnose at an early stage and is associated with poor survival. Surgical resection remains the most effective curative treatment for HCC[3,4], but the high incidence of tumor recurrence makes long-term survival after resection unsatisfactory[5,6]. Various post-resection neoadjuvant and adjuvant therapies have been developed to reduce recurrence and improve overall survival[7-11].

Transarterial chemoembolization (TACE) has recently been reported as an adjuvant therapy for HCC patients after curative resection. Originally developed as a palliative treatment to improve overall survival and reduce recurrence in patients with unresectable HCC[12,13], this treatment takes advantage of the high arterial blood supply in HCC; chemotherapy and occlusion of the feeding arteries cause tumor necrosis and delay tumor progression. TACE has been shown to provide survival benefit as a postoperative adjuvant therapy in some studies, but not in others[14-19].

To gain a clearer picture of the influence of postoperative adjuvant TACE on survival and recurrence after curative resection for HCC, we carried out a retrospective study on propensity score-matched patients without risk factors who were treated by resection followed by TACE or by resection alone.

MATERIALS AND METHODS

This study was approved by the Ethics Committee of the Tumor Hospital of Guangxi Medical University, and written informed consent was obtained from patients or from their parents or legal guardians if the patients were under 18 years of age.

Patients

This retrospective analysis was carried out on medical records of patients diagnosed with HCC between March 2007 and March 2010 at the Tumor Hospital of Guangxi Medical University, Nanning, China. A total of 1035 patients with HCC were enrolled in our hospital, but 432 (41.7%) were excluded because they had received treatment at other hospitals. Among the remaining 603 patients, 257 (42.6%) were excluded because they were given various post-resection neoadjuvant and adjuvant therapies, such as local ablation therapy or ethanol injection, systematic chemotherapy or sorafenib therapy. Another eight patients (1.3%) were excluded because they had been diagnosed with other malignancy before receiving initial HCC treatment. The remaining 229 (67.8%) patients with Child-Pugh A liver function were included in the analyses in this retrospective study. Of these patients, 138 (60.3%) underwent curative resection (control group) and 91 (39.7%) underwent resection followed by postoperative adjuvant TACE (TACE group).

Patients were enrolled if they satisfied the following criteria: (1) HCC diagnosis was confirmed by pathologic examination; (2) no more than three tumor nodules were present; (3) there was no invasion of the main trunk and first-order branches of the portal vein and common hepatic duct, or the main trunk of the hepatic vein and inferior vena cava; (4) there was no intra- or extra-hepatic metastasis, and without satellite nodules; (5) the primary tumor had been completely excised with a negative resection margin; (6) no residual tumor or portal tumor thromboses were detected in postsurgical imaging; and (7) elevated preoperative levels of alpha-fetoprotein (AFP) decreased to normal within 2 mo after the procedure.

Propensity score matching

There may be substantial risk that observed differences in survival or recurrence between our groups are due at least in part to baseline differences. To minimize such confounding and isolate the effects of adjuvant TACE, we used propensity score matching to balance these baseline differences and thereby simulate random group allocation[20]. Logistic regression was used to generate a binary propensity score for each patient (0 = assignment to control group, 1 = assignment to TACE group). One-to-one matching of propensity scores was performed without replacement using a caliper with a width 0.2 of the SD to generate 61 pairs comprised one control and one TACE patient (Table 1).

Table 1.

Baseline clinical characteristics of hepatocellular carcinoma patients treated by curative resection alone or by resection followed by transarterial chemoembolization

Variable	Before propensity score matching			After propensity score matching
	TACE (n = 91)	Control (n = 138)	P value	TACE (n = 61)	Control (n = 61)	P value
Gender
Male	79	120	0.975	50	51	0.810
Female	12	18		11	10
Age, yr	49.8 ± 10.0	49.0 ± 12.5	0.585	50.1 ± 11.0	50.8 ± 11.0	0.693
BMI, kg/m2
< 23	48	99	0.003	41	42	0.846
≥ 23	43	39		20	19
HbsAg
Positive	76	119	0.572	50	52	0.625
Negative	15	19		11	9
Family history of HCC
Yes	21	29	0.712	13	12	0.823
No	70	109		48	49
Total bilirubin, μmol/L	14.1 ± 6.9	14.0 ± 6.7	0.950	13.4 ± 5.9	14.2 ± 7.4	0.451
Total serum protein, g/L	69.1 ± 6.0	69.7 ± 6.6	0.510	69.7 ± 6.2	70.0 ± 6.3	0.783
Albumin, g/L	41.2 ± 3.9	39.9 ± 5.1	0.042	40.1 ± 4.2	40.9 ± 4.2	0.675
ALT, U/L	38 (10-192)	35 (9-341)	0.572	38 (24-53)	33 (23-49)	0.536
AST, U/L	35 (16-181)	38 (17-410)	0.507	36 (30-50)	36 (29-51)	0.903
Prothrombin time, s	12.5 ± 1.6	13.1 ± 1.7	0.010	12.6 ± 1.4	12.6 ± 1.5	0.716
Platelet count, 109/L	181.1 ± 69.2	173.2 ± 78.8	0.436	183.8 ± 73.1	174.2 ± 70.8	0.401
AFP, ng/mL
< 400	73	98	0.117	42	45	0.548
≥ 400	18	40		19	16
Tumor size, cm	5.6 ± 3.0	6.2 ± 2.9	0.155	5.8 ± 2.8	5.8 ± 2.9	0.932
Tumor number
1	67	99	0.754	44	46	0.681
2 or 3	24	39		17	15
Liver cirrhosis
Positive	78	102	0.033	51	51	1.000
Negative	13	36		10	10
Antiviral therapies
Yes	76	119	0.572	50	52	0.625
No	15	19		11	9
Resection
Anatomic	39	60	0.926	26	25	0.854
Nonanatomic	52	78		35	36
Vascular occlusion
No	22	23	0.184	14	11	0.111
Pringle maneuver	42	80		29	40
Hemihepatic	27	35		18	10
Surgical margins, cm
≥ 1	65	106	0.359	44	41	0.691
< 1	26	32		17	19
Blood loss, mL	300 (200-500)	300 (200-500)	0.119	300 (200-500)	300 (175-500)	0.726
Blood transfusion
Yes	17	38	0.125	12	13	0.823
No	74	100		49	48
Tumor capsule
Complete	59	63	0.004	36	35	0.854
Incomplete	32	75		25	26
Edmondson grade
I or II	37	60	0.673	26	28	0.715
III or IV	54	78		35	33
BCLC stage
A	68	100	0.705	46	47	0.832
B	23	38		15	14

Data are presented as mean ± SD or median (interquartile range). AFP: Alpha-fetoprotein; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; BCLC: Barcelona Clinic Liver Cancer staging system; BMI: Body mass index; HCC: Hepatocellular carcinoma; HbsAg: Hepatitis B surface antigen; TACE: Transarterial chemoembolization.

Treatment

All patients were treated by curative resection involving 1-2 segmentectomies (minor resection, n = 116) or > 2 segmentectomies (major resection, n = 113). Surgical techniques and perioperative management were as described[21,22].

Angiography of the celiac artery and superior mesenteric artery were performed in all patients, and TACE was performed 3-4 wk after curative resection when the tumor staining and visualization of tumor vasculature were observed by hepatic angiography[23]. If the remnant livers were stained by Lipiodol using CT or magnetic resonance imaging (MRI) after the hepatic angiography 1 mo later, TACE was performed. A highly selective 5-F catheter was used to completely occlude the feeding arteries and preserve liver function. In brief, a mixture of 50-100 mg cisplatin or oxaliplatin, 30-50 mg doxorubicin, and 5-10 mL of Lipiodol were injected, after which embolization was performed under fluoroscopic guidance using Gelfoam powder and a small amount of contrast medium.

Follow-up

Follow-up duration was calculated from the day of curative resection until either death or the last follow-up visit; survival time was defined as 36 mo for those who survived more than three years. All patients were followed up at 1 mo and then every 3 mo for the first two years, and then every 6 mo thereafter. Follow-up visits consisted of a physical examination, liver function tests, measurement of serum AFP, abdominal ultrasonography, and CT or MRI. Patients were diagnosed with recurrence when ultrasonography, dynamic CT, or MRI detected a new hepatic lesion.

Statistical analysis

Normally distributed clinical characteristics are reported as a single number for categorical data or as mean ± SD for continuous data. Data with a non-normal distribution are reported as median (interquartile range). Categorical data were compared between control and TACE groups using the χ² test, while continuous data were compared using a Student’s t test. Values for the mean, median, and categorical variants were compared between the two groups using the t, Mann-Whitney U, and χ² tests, respectively. Recurrence and survival curves were constructed using the Kaplan-Meier method and compared using Cox proportional hazard regression. Multivariate Cox proportional hazard regression was used to assess the ability of clinicopathologic variables to predict overall survival based on HRs and associated 95%CIs. All analyses were performed using SPSS 19.0 (IBM Corp., Armonk, NY, United States). All statistical tests were two-sided, and inter-group differences were considered significant when P < 0.05.

RESULTS

Differences between control and TACE patients before and after propensity score matching

Before propensity score matching, a comparison of all 229 patients revealed no significant differences in gender, age, rates of hepatitis B antigen positivity, antiviral therapies, family history of HCC, levels of total bilirubin, total serum protein, alanine aminotransferase, aspartate aminotransferase, or AFP, tumor size, tumor number, liver resection surgery, vascular occlusion, surgical margins, blood loss, blood transfusion, Edmondson grade, or Barcelona Clinic Liver Cancer (BCLC) stage (Table 1). However, TACE patients had significantly higher levels of serum albumin (P = 0.042), higher rates of liver cirrhosis (P = 0.033), and completely encapsulated tumors (P = 0.004). On the other hand, the control group showed significantly higher rates of body mass index < 23 kg/m² (P = 0.003) and major resection, as well as significantly longer mean prothrombin time (P = 0.010).

Propensity score matching to generate 61 patient pairs eliminated all significant differences in measured variables, confirming the effectiveness of the matching procedure.

Overall survival

Among all patients without propensity score matching, median follow-up was 32.4 mo for the TACE group and 28.7 mo for the control group. During follow-up, 51 (56.0%) TACE patients and 79 (57.2%) control patients died of cancer or complications associated with liver disease. TACE patients showed significantly higher overall survival than controls at 1 year (93.4% vs 84.1%), 2 years (86.4% vs 71.9%), and 3 years (79.2% vs 63.7%, P = 0.012; Figure 1).

Figure 1.

Kaplan-Meier curves of overall survival for hepatocellular carcinoma patients treated by curative resection alone or by resection followed by transarterial chemoembolization. Separate curves are shown for the entire patient population (P = 0.012) and for propensity score-matched patients (P = 0.543). TACE: Transarterial chemoembolization.

Median follow-up for the 61 pairs of propensity score-matched patients was 32.1 mo for TACE patients and 28.3 mo for controls. A total of 16 (26.2%) TACE patients died, compared to 25 (41.0%) control patients. Overall survival was higher among the propensity score-matched TACE patients compared to controls at 1 year (95.1% vs 86.9%), 2 years (86.7% vs 78.5%), and 3 years (76.4% vs 73.2%; Figure 1), though the difference was not significant.

Prognostic factors for overall survival

Numerous factors were assessed for their ability to predict overall survival, including age, gender, body mass index, hepatitis B infection, antiviral therapies; family history of HCC, serum biochemistry, tumor number and size, liver cirrhosis, tumor encapsulation, liver resection surgery, vascular occlusion, surgical margins, blood loss, blood transfusion, Edmondson grade, BCLC stage, resection type, and use of adjuvant TACE. Univariate analysis identified the following significant predictors of increased risk of mortality: serum AFP level ≥ 400 ng/mL (P = 0.007), tumor size > 5 cm (P < 0.001), BCLC B stage (P = 0.020), and resection without TACE (P = 0.012) (Table 2). Multivariate Cox regression identified the following independent prognostic factors: AFP level ≥ 400 ng/mL (HR = 1.745, 95%CI: 1.008-3.019; P = 0.047) and tumor size > 5 cm (HR = 2.990, 95%CI: 1.545-5.788; P = 0.001).

Table 2.

Univariate analyses to identify predictors of overall survival among all patients in the cohort

Variable	n	Overall survival (%)			Log-rank P value
		1-yr	2-yr	3-yr
AFP, ng/mL
< 400	171	89.4	80.8	74.7	0.007
≥ 400	58	82.6	68.6	55.7
Tumor size, cm
≤ 5	93	96.8	92.1	86.0	< 0.001
> 5	136	81.5	67.8	58.9
BCLC stage
A	168	89.3	80.1	74.2	0.020
B	61	83.4	70.7	59.1
Blood transfusion
Yes	55	76.4	66.7	54.9	0.030
No	174	91.3	81.1	74.3
Adjuvant TACE
Yes	91	93.4	86.4	79.2	0.012
No	138	84.1	71.9	63.7

AFP: Alpha-fetoprotein; BCLC: Barcelona Clinic Liver Cancer staging system; TACE: Transarterial chemoembolization.

Recurrence

Among all patients, recurrence rates were not significantly different in the TACE group compared to controls at 1 year (13.6% vs 24.3%), 2 years (30.4% vs 32.0%), or 3 years (40.8% vs 33.9%). Similar results were obtained for the propensity score-matched patients at 1 year (13.4% vs 24.8%), 2 years (30.6% vs 32.1%), and 3 years (40.1% vs 34.0%) (Figure 2).

Figure 2.

Kaplan-Meier curves of recurrence for hepatocellular carcinoma patients treated by curative resection alone or by resection followed by transarterial chemoembolization. Separate curves are shown for the entire patient population (P = 0.226) and for propensity score-matched patients (P = 0.425). TACE: Transarterial chemoembolization.

Prognostic factors for recurrence

Univariate analysis identified the following predictors of increased risk of recurrence: tumor size > 5 cm (P = 0.046), multiple tumors (P = 0.007), BCLC B stage (P = 0.008), and resection without TACE (P = 0.030). Multivariate Cox regression identified tumor size > 5 cm (HR = 1.684, 95%CI: 1.018-2.786; P = 0.042) as an independent prognostic factor.

DISCUSSION

The rationale for postoperative TACE is to eliminate tumor cells that may have been shed from tumor masses removed during liver surgery, as well as destroy small intrahepatic metastases that may not have been detected before or during surgery. Studies have given conflicting results concerning whether postoperative adjuvant TACE can improve patient prognosis. Our retrospective study of a reasonably large patient cohort, some of whom received resection followed by TACE and others who received resection alone, suggests that the technique does not improve overall survival or reduce recurrence.

We performed analyses not only on the entire patient cohort, but also on propensity-score matched patient pairs, as the control and TACE patients showed significant differences in some baseline characteristics. These differences reflect the criteria used in our clinic to decide whether to perform TACE on given patients; such differences have been associated with differences in overall survival in HCC[24]. Although higher one-, two-, and three-year overall survival rates were observed when all TACE patients were compared with all controls, similar to those of several studies[10,14,15], this difference was not significant when compared between score-matched patient pairs.

Our results were obtained with a cohort of HCC patients who did not have complications known to increase risk of postoperative recurrence and reduce overall survival, such as vascular invasion, satellite nodules, distant and lymphovascular metastases, and positive resection margins. One meta-analysis and two randomized controlled trials (RCTs) provided strong evidence that postoperative TACE can decrease tumor recurrence and improve overall survival in patients with such risk factors for recurrence[16,17,19,23]. It may be that postoperative TACE is more appropriate for HCC patients with numerous risk factors than for those with less complicated HCC, such as the ones in our study.

Our finding that postoperative TACE does not reduce recurrence may have several explanations. TACE reduces remnant liver function and immune function, resulting in poor overall or disease-free survival[15]. Tumors in recurrence usually have different clonal origins from the primary tumors, and liver cirrhosis and liver inflammation, both of which occurred at similar rates in our propensity score-matched patients, are independently associated with recurrence[25]. Our results suggest that applying strict selection criteria when initially deciding on HCC treatments may be more effective than postoperative adjuvant TACE at improving overall survival and reducing recurrence.

Using multivariate Cox regression, we found that a serum AFP level ≥ 400 ng/mL independently predicted poor overall survival, consistent with a previous study[26]. AFP is widely used to predict intra- and extrahepatic recurrence, with extremely high levels regarded as a highly specific and sensitive biomarker. Lower AFP levels are associated with slower tumor growth and, therefore, with less-advanced tumor stage and lower risk of mortality[27-29]. Lower AFP is also associated with a more highly differentiated HCC, and with lower risk of microscopic vascular involvement. In fact, AFP serves as an adjunct diagnostic tool in many guidelines[30].

We also identified tumor size > 5 cm as an independent predictor of survival in HCC patients, again consistent with many studies[5,6,26]. Tumors > 5 cm are associated with increased invasiveness, reflected in higher incidence of intrahepatic metastasis and portal venous invasion[31,32]. Large tumors are also associated with the presence of satellite lesions and vascular invasion, which favor tumor recurrence and poor survival[29,33,34].

In conclusion, postoperative adjuvant TACE does not appear to improve overall survival or reduce recurrence in patients lacking many of the risk factors associated with poor prognosis. However, this is a retrospective analysis that is likely subject to subtle selection biases, even after propensity score matching. Larger RCTs and RCTs involving patients with more complicated HCC should examine in more detail the effects of postoperative TACE on HCC prognosis.

ACKNOWLEDGMENTS

The authors thank Armando Chapin Rodríguez, PhD, for his language editing, which substantially improved the quality of the manuscript.

COMMENTS

Background

Clinicians already use postoperative adjuvant transarterial chemoembolization (TACE) to prevent recurrence of hepatocellular carcinoma (HCC), but its efficacy remains controversial.

Research frontiers

The high incidence of tumor recurrence makes long-term survival after resection unsatisfactory. It is important for surgeons to find post-resection neoadjuvant and adjuvant therapies in order to hinder the development of HCC after curative resection.

Innovations and breakthroughs

In this study, the authors used propensity score matching to balance baseline differences between groups, and thereby simulate random group allocation. We then analyzed the overall survival and recurrence in HCC patients who did or did not receive adjuvant TACE after curative resection. This study will provide more evidence for TACE after curative resection of HCC in the future.

Applications

The study results suggest that postoperative adjuvant TACE is not useful for HCC patients without the risk factors, such as vascular invasion, satellite nodules, distant and lymphovascular metastases, and positive resection margins.

Terminology

TACE is used to improve overall survival and reduce recurrence in patients with unresectable HCC. However, it was reported as an adjuvant therapy for HCC patients after curative resection.

Peer-review

This is an interesting study in which the authors show that postoperative adjuvant TACE does not improve overall survival or reduce recurrence in these patients lacking most of the risk factors associated with poor prognosis.